Thread by @snaidunl

Given synthetic Karl Marx, a thread on the real deal: a NLP analysis of Marx's influence in the 19th century by my amazing student @jpowerj, who is coincidentally looking for a job. Polisci depts looking for political theory + methods, look at Jeff! cs.stanford.edu/~jjacobs3/#dis… 1/N

Jeff's methodological chapter draws links between representations of text used in NLP models and the linguistic and discursive analysis of political theory pioneered by the "Cambridge school", folks like Quentin Skinner, Pocock, and Geuss.

Both Cambridge school and computational linguistics have their origins in the late, anti-positivist Wittgenstein, in their emphasis on language being used to *do* things. e.g. persuade people and win political arguments.

NLP tools like topic modelling and neural embeddings in fact capture important features of text highlighted by the Cambridge school: the need to understand writers in a context of a debate with other writers, and the need to understand a word in a context of other words.

Indeed, to understand the socialist theorizing of the 19th century, computers are essential: the writers are transnational and engaged with a mass public: without a machine to help digest the enormous amount of text only special talents could digest the relevant discursive field.

The next two chapters then apply these tools to a) understanding the influences on Marx, and b) Marx's influence on the socialist movement. He does this by reading and digitizing basically every text referenced by Marx in the MEGA, and calculating embedding distances.

Chapter 2 quantifies the "trinity" of influences:1) German philosophy, 2) French socialism, and 3) British Political Economy. Basically Marx starts out at 1), but after 1848 winds up oscillating between 2) and 3) depending on audience and topic, with 1) slowly creeping back in.

Chapter 3 then looks at the effects Marx has on European Socialism. Basically the argument is that Marx pulls European socialists away from more moralistic language, and towards the language of "scientific socialism".

Jeff does this with semantic influence networks (basically A influences B if B's current writing looks more like A's past writing than vice versa). Marx is more central in this network than his rivals in the left, Proudhon, Bakunin, and Lasalle.

Basically Marx is fighting it out with these other intellectuals for status as preeminent intellectual on the left, and wins by outposting everyone else on Twitter. Look at Jeff's dissertation for more details! N/N
academiccommons.columbia.edu/doi/10.7916/k9…